Date of Award

Spring 2018

Degree Type



Computer Science


Dr. Valerie Summet

Second Advisor

Dr. Daniel Myers

Third Advisor

Dr. Wenxian Zhang


The Rollins College Archives are a treasure trove of historic resources relating to the college’s history, and they are often underutilized or overlooked by both the student body and the surrounding community. In particular, historic resources are all too often excluded from work in computer science and related fields. This project aims to bridge that gap by bringing the two areas together. To that end, the goal of this project is to merge past and present by blending historic photos with input of a present day scene in order to reveal changes and juxtapositions of the same scene across eras. This research explores the possibility of accomplishing this principally through computational means. In order to achieve this, we delve into the domain of computer vision, utilizing techniques in feature detection and matching in order to ultimately blend images in novel ways. Image blending is a technique often used for the creation of unique images, or for emphasizing a contrast between two scenes through their convergence. Whether the blend is produced through masks with alpha values, seam carving, or other techniques, most implementations require a great deal of manual input, whether that entails point selection, mask generation, or setting an alpha value. In this project, we identify recognizable regions and features on a given image. We then use these to identify similar regions and features in a second image. Any matches found are then filtered, and the bad or incorrect matches are removed. The remaining matches are used to compute the difference in perspectives between the two images, and the coordinates of the matching points are used to correct the images to match in the same perspective. We explore various approaches to the problem of feature matching, including built-in library functions, as well as a region based, template-matching algorithm. We also investigate techniques in image blending, such as automatic mask generation, Laplacian pyramid blending, and various off-the-shelf tools contained within Unity. We also test the applications of our findings with regards to working with 360-degree images.